Outcomes of Persons With COVID-19 in Hospitals With and Without Standard Treatment With (Hydroxy)chloroquine

Objective To compare survival of subjects with COVID-19 treated in hospitals that either did or did not routinely treat patients with hydroxychloroquine or chloroquine. Methods We analysed data of COVID-19 patients treated in 9 hospitals in the Netherlands. Inclusion dates ranged from February 27th 2020, to May 15th, when the Dutch national guidelines no longer supported the use of (hydroxy)chloroquine. Seven hospitals routinely treated subjects with (hydroxy)chloroquine, two hospitals did not. Primary outcome was 21-day all-cause mortality. We performed a survival analysis using log-rank test and Cox-regression with adjustment for age, sex and covariates based on premorbid health, disease severity, and the use of steroids for adult respiratory distress syndrome, including dexamethasone. Results Among 1893 included subjects, 21-day mortality was 23.4% in 1552 subjects treated in hospitals that routinely prescribed (hydroxy)chloroquine, and 17.0% in 341 subjects that were treated in hospitals that did not. In the adjusted Cox-regression models this difference disappeared, with an adjusted hazard ratio of 1.17 (95%CI 0.88-1.55). When stratified by actually received treatment in individual subjects, the use of (hydroxy)chloroquine was associated with an increased 21-day mortality (HR 1.58; 95%CI 1.25-2.01) in the full model. Conclusions After adjustment for confounders, mortality was not significantly different in hospitals that routinely treated patients with (hydroxy)chloroquine, compared with hospitals that did not. We compared outcomes of hospital strategies rather than outcomes of individual patients to reduce the chance of indication bias. This study adds evidence against the use of (hydroxy)chloroquine in patients with COVID-19.


123
The spread of SARS-CoV-2, leading to the current pandemic of COVID-19, has a profound global 124 impact on daily life, morbidity and mortality. Several preliminary studies have reported that the 125 antimalarial agents hydroxychloroquine and chloroquine, or (H)CQ, alone or in combination with the 126 antibiotic azithromycin, can have a suppressive effect on the viral replication, and might decrease the 127 mortality of COVID-19 1-4 . So far, clinical studies have been hampered by (indication) bias 1,2,4 , 128 monocentre setup 2,3 , small numbers of included subjects 3 , and concerns about the verifiability of 129 data, even leading to withdrawal of a publication 4 . Side effects of (H)CQ are well-known and 130 common, and include fever and cardiac arrhythmias. Randomized controlled clinical trials (RCTs) are 131 currently being conducted to investigate the effect of (H)CQ on outcome of COVID-19. While we are 132 awaiting definite results from RCTs, cohort studies can provide quick closure of existing knowledge 133 gaps. When treatment assignment in cohort studies is based on prescriber discretion, the risk of 134 indication bias (even after covariate adjustment) remains high. However, our database of Dutch 135 hospitals contains data of subjects from hospitals that either routinely prescribed (H)CQ or did not 136 prescribe it at all, offering a unique opportunity to compare both strategies. The comparison of 137 different treatment strategies among hospitals leads to a significant reduction of (indication) bias. 138 The objective of this study was to compare the effect of hospital-wide COVID-19 treatment strategies 139 with or without routine (H)CQ use on all-cause 21-day mortality. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101 Methods 144 145 We used data from the ongoing CovidPredict Clinical Course Cohort containing over 2,000 persons 146 with COVID-19 5 , from 9 hospitals in the Netherlands, including two tertiary care hospitals. Included in 147 the database were all subjects admitted to hospital with positive SARS-CoV-2 PCR of nasopharynx, 148 throat, sputum or bronchoalveolar lavage samples, or CT-scan abnormalities that were typical for 149 COVID-19 (CO-RADS 4 and 5) 6 , without another explanation for the abnormalities than  Inclusion dates ranged from the first admitted case in the Netherlands on February 27 th 2020, to May 151 15 th , when the Dutch national guidelines no longer advised the use of (H)CQ. We excluded patients < 152 18 years and patients who were transferred to or from another hospital. Dosage of chloroquine base 153 was: loading dose of 600 mg, followed by 300 mg twice a day for a total of 5 days. Dosage of 154 hydroxychloroquine sulphate was 400 mg twice daily on the first day, followed by 200 mg twice daily 155 on days 2 to 5. Among the seven (H)CQ-hospitals, the timing of start of (H)CQ treatment differed; 156 three hospitals started at the moment of COVID-19 diagnosis, four started after diagnosis but only 157 when patients clinically deteriorated e.g., when there was an increase in respiratory rate or increase 158 in use of supplemental oxygen. The two hospitals that did not routinely treat subjects with (H)CQ 159 (i.e., the non-(H)CQ-hospitals), offered best supportive care, including oxygen therapy and 160 potentially antibiotic therapy, according to local guidelines and prescriber discretion. Participating 161 hospitals did not routinely prescribe other experimental medication (e.g., lopinavir/ritonavir, 162 remdesivir or steroids, see Table 1). Subjects who were incidentally treated with these drugs were 163 included in the study. Primary outcome was 21-day all-cause mortality, defined as hospital mortality, 164 or discharge to a hospice care facility. A waiver for the use of hospital record data was obtained 165 through the Institutional Review Board of Amsterdam UMC; however, patients were given the 166 opportunity to opt out. In the primary analysis, we compared effectiveness of (H)CQ versus non-167 (H)CQ hospital strategies, irrespective of actual individual (H)CQ treatment. We performed a survival 168 analysis using log-rank test and Cox-regression with adjustment for age, sex and covariates based on 169 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . premorbid health (i.e., history of lung, kidney and cardiovascular disease, diabetes mellitus, obesity, 170 and neoplasms or hematologic disease), disease severity during presentation (respiratory rate, 171 oxygen saturation) and the use of steroids, including dexamethasone, for adult respiratory distress 172 syndrome (ARDS) 7,8 . We collected data according to the collection protocol of the World Health 173 Organization. Missing covariates were imputed using multiple imputation. As a sensitivity analysis, 174 we performed a complete case analysis using inverse probability weighting of propensity scores 175 (determined using the same covariates), and a subgroup analysis in hospitals who started (H)CQ from 176 the moment of diagnosis. Finally, we repeated the analyses comparing actually received treatment, 177 with (H)CQ. All statistical analyses were performed using R versions 3.6.3 (R Foundation, Vienna, 178 Austria). 179 180

182
We analysed results of 1893 subjects admitted before May 15 th 2020, 239 were excluded because 183 they were transferred from another hospital. Demographic data are shown in Table 1 peripheral oxygen saturation during admission were similar in both hospital groups (see Table 1). In 193 (H)CQ-hospitals, 9.7% of subjects received corticosteroids for ARDS and 3.2% were in a study 194 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint protocol of an experimental SARS-CoV-2 directed antiviral (e.g., lopinavir/ritonavir) or 195 immunomodulatory drug trial (e.g., imatinib, anti-complement C5), versus 2. 4% and 11.3% in non-196 (H)CQ-hospitals, respectively. Figure 1 shows the survival of subjects in (H)CQ-versus non-(H)  hospitals. Unadjusted mortality at day 21 was significantly different between the (H)  (H)CQ-hospitals (23.4% vs. 17.0%). However, in the adjusted Cox-regression models, this difference 199 disappeared, with an adjusted hazard ratio of 1.17 (95%CI 0.88-1.55, Table 2). The sensitivity analysis 200 of hospitals routinely starting (H)CQ treatment from the moment of COVID-19 diagnosis (i.e., (H)CQ 201 hospitals without the hospitals that initiated (H)CQ treatment upon clinical deterioration) compared 202 with non-(H)CQ-hospitals, showed similar results with a HR of 1.14 (95%CI 0.82-1.57) after 203 adjustment for age, sex, comorbidities, and disease severity at presentation. Complete case analysis 204 and the analysis using propensity scores showed similar results (see Table 3). Finally, when stratified 205 by actually received treatment, the use of (H)CQ was associated with an increased 21-day mortality 206 (HR 1.58; 95%CI 1.25-2.01, Table 3) in the full model. The sensitivity analysis of all hospitals and all 207 subjects can be found in Table 4. 208 209 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint At first glance, our study results may seem to suggest that subjects treated in hospitals that routinely 215 prescribed (H)CQ had a significantly increased 21-day all-cause mortality compared with those in 216 hospitals that did not routinely prescribe (H)CQ. However, mortality was not significantly different 217 after adjustment for age, sex, medical history, disease severity at presentation and steroid use during 218 treatment. Similarly, we found an increased risk of death among subjects who had actually received 219 treatment with (H)CQ, which has likely been driven by indication bias, as in four of the seven (H)CQ-220 hospitals, (H)CQ was only prescribed upon clinical deterioration. The unique characteristics of our 221 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint study cohort enabled a study design that minimized indication bias. Our results add further weight to 222 existing evidence against the use of (H)CQ for the treatment  224 The strength of this study is that data were collected in nine secondary and tertiary care hospitals in 225 the Netherlands during the COVID-19 epidemic. Data collection was set up prospectively and the 226 database included data on all consecutive subjects admitted to general medicine and pulmonology 227 wards, and to intensive care units. The database was set up according to the WHO standards, which 228 enabled data comparison and uniformity of data among the different participating centres. The 229 comparison of hospital-defined treatment strategies rather than the treatment actually received led 230 to a lower risk of indication bias compared with previous studies 1,2 . We roughly estimate the extend 231 of the effect of indication bias to be the difference in outcome between the uncorrected and the 232 corrected model. Further strengths include the multicentre setup 2,3 , as mentioned above, and the 233 relatively large numbers of included subjects 3 . 234

235
There are some limitations we need to address. Although health care in the Netherlands has a 236 homogeneous setup, there was some variability in standard protocols among the hospitals that could 237 have led to residual confounding. The two non-(H)CQ-hospitals were tertiary (academic) centres, 238 whereas the (H)CQ-hospitals comprised secondary care hospitals. Since we excluded subjects 239 transferred to and from other hospitals, the referral role of the tertiary care hospitals was minimized. 240 Furthermore, subjects in the (H)CQ hospitals were more likely to receive steroid treatment, while 241 subjects in the non-(H)CQ hospitals were more likely to receive other experimental 242 immunomodulatory drugs. The numbers of the individual types of medication were small, making it 243 impossible to draw conclusions from these differences. The results of the RECOVERY trial (publication 244 pending), suggested a lower mortality in patients treated with dexamethasone. Treatment with 245 dexamethasone could therefore have resulted in a lower mortality in the group of (H)CQ hospitals. 246 We did not find such an effect, even after correction in the full model. We also used extensive 247 . CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . covariate adjustments, using various methods to minimize influence of differences in patient 248 population among hospitals, and the similarity in outcomes between these methods is reassuring in 249 this regard. Finally, because not every subject in the (H)CQ-hospitals actually received (H)CQ, the 250 current efficacy estimate in our study is likely an underestimation of the true (H)CQ effect. 251 Performing an instrumental variable analysis would have provided an approximation of this true 252 effect, but because the current efficacy point estimates point toward harm rather than benefit of 253 (H)CQ, this likely would not have changed our conclusions. 9 254 255 Despite the positive results of some studies resulting in widespread use of (H)CQ, our study did not 256 show a benefit of (H)CQ treatment. This may be explained by the timing of the administration of the 257 drug and its specific working mechanism. Chloroquine binds in silico and in vitro with high affinity to 258 sialic acids and gangliosides of SARS-CoV-2. These bindings inhibit the interaction at non-toxic plasma 259 levels with ACE-2 receptors and could hypothetically stop the cascade from formation of pulmonary 260 infiltrations to full blown ARDS and death 10-12 . The antiviral activity might be more effective in the 261 pre-clinical setting as the deterioration in the hospital is more an effect of the cytokine storm 262 provoked by SARS-CoV-2 than an effect of the viral infection itself. This hypothesis might explain why 263 the clinical benefit for admitted subjects was absent in our study, although we did not observe a 264 difference in outcome among subjects treated early (at diagnosis) and among those treated later 265 upon clinical deterioration. Currently, clinical trials are underway to study (hydroxy)chloroquine in 266 admitted COVID-19 patients (e.g., NCT04261517 and NCT04307693). As post-exposure prophylaxis, 267 hydroxychloroquine does not seem to be effective to prevent COVID-19, 13 but results of further 268 studies are pending 14 . Given the current evidence, we would argue against the use (H)CQ outside the 269 setting of randomized clinical trials. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. .

submitted. 290
. CC-BY 4.0 International license It is made available under a perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101  is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted August 15, 2020. . https://doi.org/10.1101/2020.08.14.20173369 doi: medRxiv preprint